Abstract: Owing to the large-scale integration of flexible load resources, the distribution network has become increasingly complex and active. The frequent variation in the topology of distribution networks poses great challenges to the operation, analysis and control of distribution networks. To address the limitation of existing global topology identification methods which have not adequately considered regional differences, a partition topology identification method is proposed based on grid SVM for active distribution networks. Firstly, an active distribution network global topology identification model is established based on SVM, and the real-time measurement data of distribution network are preliminarily identified. Subsequently, the electrical distance between nodes is calculated to mesh the distribution network topology obtained by the preliminary identification. A topology identification model based on grid SVM is then established to identify the partition topology of the distribution network, and detect and correct the initial topology identification errors. Finally, to further enhance the accuracy of topology identification, the cyclic iterative method is utilized to repeatedly performing meshing and partition topology identification validation, so as to obtain the best topology identification result. The superiority of the proposed method is verified through an instance of an IEEE 33-bus distribution network.